Textual Information in Analyst Reports and its Value for Decision Support
by Matthias Palmer
Date of Examination:2022-11-17
Date of issue:2023-02-16
Advisor:Prof. Dr. Jan Muntermann
Referee:Prof. Dr. Jan Muntermann
Referee:Prof. Dr. Matthias Schumann
Referee:Prof. Dr. Lutz M. Kolbe
Files in this item
This file will be freely accessible after 2024-06-01.
EnglishThe profession of financial analysts requires skills that cannot be replaced even by today’s most cutting-edge analytical tools. Instead, ongoing globalization and the increasing interconnectedness of international financial markets are creating new levels of complexity that did not exist even a few decades ago. Therefore, this thesis is dedicated to the improved processing and use of textual analyst opinion and promoting the value of this data for decision support. The intention is to take a pragmatic and user-oriented view, which remained in the background of previous research on fundamental relationships. Specifically, this thesis aims to successfully apply the signals from analyst reports in actual use cases. Here, special attention is paid to the settings in which analyst reports should be read with caution and which analyst-specific characteristics should be considered in the decision-making process. In light of the outlined objectives, the thesis is divided into three research areas: methodological foundations, behavioral patterns of financial analysts, and analyst reports and financial markets. The first two research areas serve as a basis. They cover selected approaches in text mining on the one hand and typical behavioral patterns visible in qualitative analyst content on the other hand. The first research area, methodological foundations, proposes a framework that includes different approaches to document representation, providing a guideline for conducting text mining tasks in various domains. Additionally, a sentiment dictionary is developed that improves the domain-specific analysis of analyst content. The second research area, behavioral patterns of financial analysts, addresses the detection of herding behavior among financial analysts by applying topic mining. Furthermore, a particularly pronounced linguistic distinction between analyst reports and other finance-specific texts is illustrated. The third research area, analyst reports and financial markets, builds on the previous two areas and displays the practical application of textual analyst reports by showcasing the implementation of an investment strategy and a risk management approach based on textual analyst reports. Overall, the three research areas in this thesis span from considering methodological aspects and understanding the distinctive characteristics of texts in analyst reports to demonstrating relevant practical applications. In combination, the individual studies aim to contribute to a better understanding of the potential of qualitative analyst information and emphasize the role of this information as a sensible component of decision support systems.
Keywords: Financial Analysts, Analyst Reports, Unstructured Data, Text Mining, Sentiment Analysis, Sentiment Dictionary, Topic Mining, Decision Support, Portfolio Strategy, Credit Risk